A comparative study of discrete, semicontinuous, and continuous hidden Markov models
In this paper, we first extended the semicontinuous hidden Markov model to incorporate multiple code-books. The robustness of the semicontinuous output probability is enhanced by the combination of multiple codewords and multiple codebooks. In addition, we compared the semicontinuous model with the...
Ausführliche Beschreibung
Gespeichert in:
| Autor*in: |
|---|
| Format: |
E-Artikel |
|---|---|
| Sprache: |
Englisch |
| Erschienen: |
1993 |
|---|
| Reproduktion: |
Elsevier Journal Backfiles on ScienceDirect 1907 - 2002 |
|---|---|
| Übergeordnetes Werk: |
in: Computer Speech & Language - Amsterdam : Elsevier, 7(1993), 4, Seite 359-368 |
| Übergeordnetes Werk: |
volume:7 ; year:1993 ; number:4 ; pages:359-368 |
| Links: |
|---|
| Katalog-ID: |
NLEJ184522943 |
|---|
| LEADER | 01000caa a22002652 4500 | ||
|---|---|---|---|
| 001 | NLEJ184522943 | ||
| 003 | DE-627 | ||
| 005 | 20210706232749.0 | ||
| 007 | cr uuu---uuuuu | ||
| 008 | 070506s1993 xx |||||o 00| ||eng c | ||
| 035 | |a (DE-627)NLEJ184522943 | ||
| 035 | |a (DE-599)GBVNLZ184522943 | ||
| 040 | |a DE-627 |b ger |c DE-627 |e rakwb | ||
| 041 | |a eng | ||
| 245 | 1 | 2 | |a A comparative study of discrete, semicontinuous, and continuous hidden Markov models |
| 264 | 1 | |c 1993 | |
| 336 | |a nicht spezifiziert |b zzz |2 rdacontent | ||
| 337 | |a nicht spezifiziert |b z |2 rdamedia | ||
| 338 | |a nicht spezifiziert |b zu |2 rdacarrier | ||
| 520 | |a In this paper, we first extended the semicontinuous hidden Markov model to incorporate multiple code-books. The robustness of the semicontinuous output probability is enhanced by the combination of multiple codewords and multiple codebooks. In addition, we compared the semicontinuous model with the continuous mixture model and the discrete model in a large-vocabulary speaker-independent continuous speech recognition (DARPA resource management) task. The model assumption and parameter size issues are addressed in particular through these experiments. When the acoustic parameters are not well modelled by the continuous probability density, the model assumption problems may cause the recognition accuracy of the semicontinuous model or the continuous mixture model to be inferior to the discrete model. We also found that the SCHMM can have a large number of free parameters in comparison with the discrete HMM because of its smoothing ability. With explicit male and female clustered models and for conditional feature sets, we were able to reduce the error rate of discrete-model-based SPHINX by more than 20%. | ||
| 533 | |f Elsevier Journal Backfiles on ScienceDirect 1907 - 2002 | ||
| 700 | 1 | |a Huang, X.D. |4 oth | |
| 700 | 1 | |a Hon, H.W. |4 oth | |
| 700 | 1 | |a Hwang, M.Y. |4 oth | |
| 700 | 1 | |a Lee, K.F. |4 oth | |
| 773 | 0 | 8 | |i in |t Computer Speech & Language |d Amsterdam : Elsevier |g 7(1993), 4, Seite 359-368 |w (DE-627)NLEJ177265698 |w (DE-600)1462883-1 |x 0885-2308 |7 nnns |
| 773 | 1 | 8 | |g volume:7 |g year:1993 |g number:4 |g pages:359-368 |
| 856 | 4 | 0 | |u http://dx.doi.org/10.1006/csla.1993.1019 |
| 912 | |a GBV_USEFLAG_H | ||
| 912 | |a ZDB-1-SDJ | ||
| 912 | |a GBV_NL_ARTICLE | ||
| 951 | |a AR | ||
| 952 | |d 7 |j 1993 |e 4 |h 359-368 | ||
| matchkey_str |
article:08852308:1993----::cmaaietdodsrtsmcniuuadotnos |
|---|---|
| hierarchy_sort_str |
1993 |
| publishDate |
1993 |
| allfields |
(DE-627)NLEJ184522943 (DE-599)GBVNLZ184522943 DE-627 ger DE-627 rakwb eng A comparative study of discrete, semicontinuous, and continuous hidden Markov models 1993 nicht spezifiziert zzz rdacontent nicht spezifiziert z rdamedia nicht spezifiziert zu rdacarrier In this paper, we first extended the semicontinuous hidden Markov model to incorporate multiple code-books. The robustness of the semicontinuous output probability is enhanced by the combination of multiple codewords and multiple codebooks. In addition, we compared the semicontinuous model with the continuous mixture model and the discrete model in a large-vocabulary speaker-independent continuous speech recognition (DARPA resource management) task. The model assumption and parameter size issues are addressed in particular through these experiments. When the acoustic parameters are not well modelled by the continuous probability density, the model assumption problems may cause the recognition accuracy of the semicontinuous model or the continuous mixture model to be inferior to the discrete model. We also found that the SCHMM can have a large number of free parameters in comparison with the discrete HMM because of its smoothing ability. With explicit male and female clustered models and for conditional feature sets, we were able to reduce the error rate of discrete-model-based SPHINX by more than 20%. Elsevier Journal Backfiles on ScienceDirect 1907 - 2002 Huang, X.D. oth Hon, H.W. oth Hwang, M.Y. oth Lee, K.F. oth in Computer Speech & Language Amsterdam : Elsevier 7(1993), 4, Seite 359-368 (DE-627)NLEJ177265698 (DE-600)1462883-1 0885-2308 nnns volume:7 year:1993 number:4 pages:359-368 http://dx.doi.org/10.1006/csla.1993.1019 GBV_USEFLAG_H ZDB-1-SDJ GBV_NL_ARTICLE AR 7 1993 4 359-368 |
| spelling |
(DE-627)NLEJ184522943 (DE-599)GBVNLZ184522943 DE-627 ger DE-627 rakwb eng A comparative study of discrete, semicontinuous, and continuous hidden Markov models 1993 nicht spezifiziert zzz rdacontent nicht spezifiziert z rdamedia nicht spezifiziert zu rdacarrier In this paper, we first extended the semicontinuous hidden Markov model to incorporate multiple code-books. The robustness of the semicontinuous output probability is enhanced by the combination of multiple codewords and multiple codebooks. In addition, we compared the semicontinuous model with the continuous mixture model and the discrete model in a large-vocabulary speaker-independent continuous speech recognition (DARPA resource management) task. The model assumption and parameter size issues are addressed in particular through these experiments. When the acoustic parameters are not well modelled by the continuous probability density, the model assumption problems may cause the recognition accuracy of the semicontinuous model or the continuous mixture model to be inferior to the discrete model. We also found that the SCHMM can have a large number of free parameters in comparison with the discrete HMM because of its smoothing ability. With explicit male and female clustered models and for conditional feature sets, we were able to reduce the error rate of discrete-model-based SPHINX by more than 20%. Elsevier Journal Backfiles on ScienceDirect 1907 - 2002 Huang, X.D. oth Hon, H.W. oth Hwang, M.Y. oth Lee, K.F. oth in Computer Speech & Language Amsterdam : Elsevier 7(1993), 4, Seite 359-368 (DE-627)NLEJ177265698 (DE-600)1462883-1 0885-2308 nnns volume:7 year:1993 number:4 pages:359-368 http://dx.doi.org/10.1006/csla.1993.1019 GBV_USEFLAG_H ZDB-1-SDJ GBV_NL_ARTICLE AR 7 1993 4 359-368 |
| allfields_unstemmed |
(DE-627)NLEJ184522943 (DE-599)GBVNLZ184522943 DE-627 ger DE-627 rakwb eng A comparative study of discrete, semicontinuous, and continuous hidden Markov models 1993 nicht spezifiziert zzz rdacontent nicht spezifiziert z rdamedia nicht spezifiziert zu rdacarrier In this paper, we first extended the semicontinuous hidden Markov model to incorporate multiple code-books. The robustness of the semicontinuous output probability is enhanced by the combination of multiple codewords and multiple codebooks. In addition, we compared the semicontinuous model with the continuous mixture model and the discrete model in a large-vocabulary speaker-independent continuous speech recognition (DARPA resource management) task. The model assumption and parameter size issues are addressed in particular through these experiments. When the acoustic parameters are not well modelled by the continuous probability density, the model assumption problems may cause the recognition accuracy of the semicontinuous model or the continuous mixture model to be inferior to the discrete model. We also found that the SCHMM can have a large number of free parameters in comparison with the discrete HMM because of its smoothing ability. With explicit male and female clustered models and for conditional feature sets, we were able to reduce the error rate of discrete-model-based SPHINX by more than 20%. Elsevier Journal Backfiles on ScienceDirect 1907 - 2002 Huang, X.D. oth Hon, H.W. oth Hwang, M.Y. oth Lee, K.F. oth in Computer Speech & Language Amsterdam : Elsevier 7(1993), 4, Seite 359-368 (DE-627)NLEJ177265698 (DE-600)1462883-1 0885-2308 nnns volume:7 year:1993 number:4 pages:359-368 http://dx.doi.org/10.1006/csla.1993.1019 GBV_USEFLAG_H ZDB-1-SDJ GBV_NL_ARTICLE AR 7 1993 4 359-368 |
| allfieldsGer |
(DE-627)NLEJ184522943 (DE-599)GBVNLZ184522943 DE-627 ger DE-627 rakwb eng A comparative study of discrete, semicontinuous, and continuous hidden Markov models 1993 nicht spezifiziert zzz rdacontent nicht spezifiziert z rdamedia nicht spezifiziert zu rdacarrier In this paper, we first extended the semicontinuous hidden Markov model to incorporate multiple code-books. The robustness of the semicontinuous output probability is enhanced by the combination of multiple codewords and multiple codebooks. In addition, we compared the semicontinuous model with the continuous mixture model and the discrete model in a large-vocabulary speaker-independent continuous speech recognition (DARPA resource management) task. The model assumption and parameter size issues are addressed in particular through these experiments. When the acoustic parameters are not well modelled by the continuous probability density, the model assumption problems may cause the recognition accuracy of the semicontinuous model or the continuous mixture model to be inferior to the discrete model. We also found that the SCHMM can have a large number of free parameters in comparison with the discrete HMM because of its smoothing ability. With explicit male and female clustered models and for conditional feature sets, we were able to reduce the error rate of discrete-model-based SPHINX by more than 20%. Elsevier Journal Backfiles on ScienceDirect 1907 - 2002 Huang, X.D. oth Hon, H.W. oth Hwang, M.Y. oth Lee, K.F. oth in Computer Speech & Language Amsterdam : Elsevier 7(1993), 4, Seite 359-368 (DE-627)NLEJ177265698 (DE-600)1462883-1 0885-2308 nnns volume:7 year:1993 number:4 pages:359-368 http://dx.doi.org/10.1006/csla.1993.1019 GBV_USEFLAG_H ZDB-1-SDJ GBV_NL_ARTICLE AR 7 1993 4 359-368 |
| allfieldsSound |
(DE-627)NLEJ184522943 (DE-599)GBVNLZ184522943 DE-627 ger DE-627 rakwb eng A comparative study of discrete, semicontinuous, and continuous hidden Markov models 1993 nicht spezifiziert zzz rdacontent nicht spezifiziert z rdamedia nicht spezifiziert zu rdacarrier In this paper, we first extended the semicontinuous hidden Markov model to incorporate multiple code-books. The robustness of the semicontinuous output probability is enhanced by the combination of multiple codewords and multiple codebooks. In addition, we compared the semicontinuous model with the continuous mixture model and the discrete model in a large-vocabulary speaker-independent continuous speech recognition (DARPA resource management) task. The model assumption and parameter size issues are addressed in particular through these experiments. When the acoustic parameters are not well modelled by the continuous probability density, the model assumption problems may cause the recognition accuracy of the semicontinuous model or the continuous mixture model to be inferior to the discrete model. We also found that the SCHMM can have a large number of free parameters in comparison with the discrete HMM because of its smoothing ability. With explicit male and female clustered models and for conditional feature sets, we were able to reduce the error rate of discrete-model-based SPHINX by more than 20%. Elsevier Journal Backfiles on ScienceDirect 1907 - 2002 Huang, X.D. oth Hon, H.W. oth Hwang, M.Y. oth Lee, K.F. oth in Computer Speech & Language Amsterdam : Elsevier 7(1993), 4, Seite 359-368 (DE-627)NLEJ177265698 (DE-600)1462883-1 0885-2308 nnns volume:7 year:1993 number:4 pages:359-368 http://dx.doi.org/10.1006/csla.1993.1019 GBV_USEFLAG_H ZDB-1-SDJ GBV_NL_ARTICLE AR 7 1993 4 359-368 |
| language |
English |
| source |
in Computer Speech & Language 7(1993), 4, Seite 359-368 volume:7 year:1993 number:4 pages:359-368 |
| sourceStr |
in Computer Speech & Language 7(1993), 4, Seite 359-368 volume:7 year:1993 number:4 pages:359-368 |
| format_phy_str_mv |
Article |
| institution |
findex.gbv.de |
| isfreeaccess_bool |
false |
| container_title |
Computer Speech & Language |
| authorswithroles_txt_mv |
Huang, X.D. @@oth@@ Hon, H.W. @@oth@@ Hwang, M.Y. @@oth@@ Lee, K.F. @@oth@@ |
| publishDateDaySort_date |
1993-01-01T00:00:00Z |
| hierarchy_top_id |
NLEJ177265698 |
| id |
NLEJ184522943 |
| language_de |
englisch |
| fullrecord |
<?xml version="1.0" encoding="UTF-8"?><collection xmlns="http://www.loc.gov/MARC21/slim"><record><leader>01000caa a22002652 4500</leader><controlfield tag="001">NLEJ184522943</controlfield><controlfield tag="003">DE-627</controlfield><controlfield tag="005">20210706232749.0</controlfield><controlfield tag="007">cr uuu---uuuuu</controlfield><controlfield tag="008">070506s1993 xx |||||o 00| ||eng c</controlfield><datafield tag="035" ind1=" " ind2=" "><subfield code="a">(DE-627)NLEJ184522943</subfield></datafield><datafield tag="035" ind1=" " ind2=" "><subfield code="a">(DE-599)GBVNLZ184522943</subfield></datafield><datafield tag="040" ind1=" " ind2=" "><subfield code="a">DE-627</subfield><subfield code="b">ger</subfield><subfield code="c">DE-627</subfield><subfield code="e">rakwb</subfield></datafield><datafield tag="041" ind1=" " ind2=" "><subfield code="a">eng</subfield></datafield><datafield tag="245" ind1="1" ind2="2"><subfield code="a">A comparative study of discrete, semicontinuous, and continuous hidden Markov models</subfield></datafield><datafield tag="264" ind1=" " ind2="1"><subfield code="c">1993</subfield></datafield><datafield tag="336" ind1=" " ind2=" "><subfield code="a">nicht spezifiziert</subfield><subfield code="b">zzz</subfield><subfield code="2">rdacontent</subfield></datafield><datafield tag="337" ind1=" " ind2=" "><subfield code="a">nicht spezifiziert</subfield><subfield code="b">z</subfield><subfield code="2">rdamedia</subfield></datafield><datafield tag="338" ind1=" " ind2=" "><subfield code="a">nicht spezifiziert</subfield><subfield code="b">zu</subfield><subfield code="2">rdacarrier</subfield></datafield><datafield tag="520" ind1=" " ind2=" "><subfield code="a">In this paper, we first extended the semicontinuous hidden Markov model to incorporate multiple code-books. The robustness of the semicontinuous output probability is enhanced by the combination of multiple codewords and multiple codebooks. In addition, we compared the semicontinuous model with the continuous mixture model and the discrete model in a large-vocabulary speaker-independent continuous speech recognition (DARPA resource management) task. The model assumption and parameter size issues are addressed in particular through these experiments. When the acoustic parameters are not well modelled by the continuous probability density, the model assumption problems may cause the recognition accuracy of the semicontinuous model or the continuous mixture model to be inferior to the discrete model. We also found that the SCHMM can have a large number of free parameters in comparison with the discrete HMM because of its smoothing ability. With explicit male and female clustered models and for conditional feature sets, we were able to reduce the error rate of discrete-model-based SPHINX by more than 20%.</subfield></datafield><datafield tag="533" ind1=" " ind2=" "><subfield code="f">Elsevier Journal Backfiles on ScienceDirect 1907 - 2002</subfield></datafield><datafield tag="700" ind1="1" ind2=" "><subfield code="a">Huang, X.D.</subfield><subfield code="4">oth</subfield></datafield><datafield tag="700" ind1="1" ind2=" "><subfield code="a">Hon, H.W.</subfield><subfield code="4">oth</subfield></datafield><datafield tag="700" ind1="1" ind2=" "><subfield code="a">Hwang, M.Y.</subfield><subfield code="4">oth</subfield></datafield><datafield tag="700" ind1="1" ind2=" "><subfield code="a">Lee, K.F.</subfield><subfield code="4">oth</subfield></datafield><datafield tag="773" ind1="0" ind2="8"><subfield code="i">in</subfield><subfield code="t">Computer Speech & Language</subfield><subfield code="d">Amsterdam : Elsevier</subfield><subfield code="g">7(1993), 4, Seite 359-368</subfield><subfield code="w">(DE-627)NLEJ177265698</subfield><subfield code="w">(DE-600)1462883-1</subfield><subfield code="x">0885-2308</subfield><subfield code="7">nnns</subfield></datafield><datafield tag="773" ind1="1" ind2="8"><subfield code="g">volume:7</subfield><subfield code="g">year:1993</subfield><subfield code="g">number:4</subfield><subfield code="g">pages:359-368</subfield></datafield><datafield tag="856" ind1="4" ind2="0"><subfield code="u">http://dx.doi.org/10.1006/csla.1993.1019</subfield></datafield><datafield tag="912" ind1=" " ind2=" "><subfield code="a">GBV_USEFLAG_H</subfield></datafield><datafield tag="912" ind1=" " ind2=" "><subfield code="a">ZDB-1-SDJ</subfield></datafield><datafield tag="912" ind1=" " ind2=" "><subfield code="a">GBV_NL_ARTICLE</subfield></datafield><datafield tag="951" ind1=" " ind2=" "><subfield code="a">AR</subfield></datafield><datafield tag="952" ind1=" " ind2=" "><subfield code="d">7</subfield><subfield code="j">1993</subfield><subfield code="e">4</subfield><subfield code="h">359-368</subfield></datafield></record></collection>
|
| series2 |
Elsevier Journal Backfiles on ScienceDirect 1907 - 2002 |
| ppnlink_with_tag_str_mv |
@@773@@(DE-627)NLEJ177265698 |
| format |
electronic Article |
| delete_txt_mv |
keep |
| collection |
NL |
| remote_str |
true |
| illustrated |
Not Illustrated |
| issn |
0885-2308 |
| topic_title |
A comparative study of discrete, semicontinuous, and continuous hidden Markov models |
| format_facet |
Elektronische Aufsätze Aufsätze Elektronische Ressource |
| format_main_str_mv |
Text Zeitschrift/Artikel |
| carriertype_str_mv |
zu |
| author2_variant |
x h xh h h hh m h mh k l kl |
| hierarchy_parent_title |
Computer Speech & Language |
| hierarchy_parent_id |
NLEJ177265698 |
| hierarchy_top_title |
Computer Speech & Language |
| isfreeaccess_txt |
false |
| familylinks_str_mv |
(DE-627)NLEJ177265698 (DE-600)1462883-1 |
| title |
A comparative study of discrete, semicontinuous, and continuous hidden Markov models |
| spellingShingle |
A comparative study of discrete, semicontinuous, and continuous hidden Markov models |
| ctrlnum |
(DE-627)NLEJ184522943 (DE-599)GBVNLZ184522943 |
| title_full |
A comparative study of discrete, semicontinuous, and continuous hidden Markov models |
| journal |
Computer Speech & Language |
| journalStr |
Computer Speech & Language |
| lang_code |
eng |
| isOA_bool |
false |
| recordtype |
marc |
| publishDateSort |
1993 |
| contenttype_str_mv |
zzz |
| container_start_page |
359 |
| container_volume |
7 |
| format_se |
Elektronische Aufsätze |
| title_sort |
comparative study of discrete, semicontinuous, and continuous hidden markov models |
| title_auth |
A comparative study of discrete, semicontinuous, and continuous hidden Markov models |
| abstract |
In this paper, we first extended the semicontinuous hidden Markov model to incorporate multiple code-books. The robustness of the semicontinuous output probability is enhanced by the combination of multiple codewords and multiple codebooks. In addition, we compared the semicontinuous model with the continuous mixture model and the discrete model in a large-vocabulary speaker-independent continuous speech recognition (DARPA resource management) task. The model assumption and parameter size issues are addressed in particular through these experiments. When the acoustic parameters are not well modelled by the continuous probability density, the model assumption problems may cause the recognition accuracy of the semicontinuous model or the continuous mixture model to be inferior to the discrete model. We also found that the SCHMM can have a large number of free parameters in comparison with the discrete HMM because of its smoothing ability. With explicit male and female clustered models and for conditional feature sets, we were able to reduce the error rate of discrete-model-based SPHINX by more than 20%. |
| abstractGer |
In this paper, we first extended the semicontinuous hidden Markov model to incorporate multiple code-books. The robustness of the semicontinuous output probability is enhanced by the combination of multiple codewords and multiple codebooks. In addition, we compared the semicontinuous model with the continuous mixture model and the discrete model in a large-vocabulary speaker-independent continuous speech recognition (DARPA resource management) task. The model assumption and parameter size issues are addressed in particular through these experiments. When the acoustic parameters are not well modelled by the continuous probability density, the model assumption problems may cause the recognition accuracy of the semicontinuous model or the continuous mixture model to be inferior to the discrete model. We also found that the SCHMM can have a large number of free parameters in comparison with the discrete HMM because of its smoothing ability. With explicit male and female clustered models and for conditional feature sets, we were able to reduce the error rate of discrete-model-based SPHINX by more than 20%. |
| abstract_unstemmed |
In this paper, we first extended the semicontinuous hidden Markov model to incorporate multiple code-books. The robustness of the semicontinuous output probability is enhanced by the combination of multiple codewords and multiple codebooks. In addition, we compared the semicontinuous model with the continuous mixture model and the discrete model in a large-vocabulary speaker-independent continuous speech recognition (DARPA resource management) task. The model assumption and parameter size issues are addressed in particular through these experiments. When the acoustic parameters are not well modelled by the continuous probability density, the model assumption problems may cause the recognition accuracy of the semicontinuous model or the continuous mixture model to be inferior to the discrete model. We also found that the SCHMM can have a large number of free parameters in comparison with the discrete HMM because of its smoothing ability. With explicit male and female clustered models and for conditional feature sets, we were able to reduce the error rate of discrete-model-based SPHINX by more than 20%. |
| collection_details |
GBV_USEFLAG_H ZDB-1-SDJ GBV_NL_ARTICLE |
| container_issue |
4 |
| title_short |
A comparative study of discrete, semicontinuous, and continuous hidden Markov models |
| url |
http://dx.doi.org/10.1006/csla.1993.1019 |
| remote_bool |
true |
| author2 |
Huang, X.D. Hon, H.W. Hwang, M.Y. Lee, K.F. |
| author2Str |
Huang, X.D. Hon, H.W. Hwang, M.Y. Lee, K.F. |
| ppnlink |
NLEJ177265698 |
| mediatype_str_mv |
z |
| isOA_txt |
false |
| hochschulschrift_bool |
false |
| author2_role |
oth oth oth oth |
| up_date |
2024-07-06T01:33:58.939Z |
| _version_ |
1803791513880100864 |
| fullrecord_marcxml |
<?xml version="1.0" encoding="UTF-8"?><collection xmlns="http://www.loc.gov/MARC21/slim"><record><leader>01000caa a22002652 4500</leader><controlfield tag="001">NLEJ184522943</controlfield><controlfield tag="003">DE-627</controlfield><controlfield tag="005">20210706232749.0</controlfield><controlfield tag="007">cr uuu---uuuuu</controlfield><controlfield tag="008">070506s1993 xx |||||o 00| ||eng c</controlfield><datafield tag="035" ind1=" " ind2=" "><subfield code="a">(DE-627)NLEJ184522943</subfield></datafield><datafield tag="035" ind1=" " ind2=" "><subfield code="a">(DE-599)GBVNLZ184522943</subfield></datafield><datafield tag="040" ind1=" " ind2=" "><subfield code="a">DE-627</subfield><subfield code="b">ger</subfield><subfield code="c">DE-627</subfield><subfield code="e">rakwb</subfield></datafield><datafield tag="041" ind1=" " ind2=" "><subfield code="a">eng</subfield></datafield><datafield tag="245" ind1="1" ind2="2"><subfield code="a">A comparative study of discrete, semicontinuous, and continuous hidden Markov models</subfield></datafield><datafield tag="264" ind1=" " ind2="1"><subfield code="c">1993</subfield></datafield><datafield tag="336" ind1=" " ind2=" "><subfield code="a">nicht spezifiziert</subfield><subfield code="b">zzz</subfield><subfield code="2">rdacontent</subfield></datafield><datafield tag="337" ind1=" " ind2=" "><subfield code="a">nicht spezifiziert</subfield><subfield code="b">z</subfield><subfield code="2">rdamedia</subfield></datafield><datafield tag="338" ind1=" " ind2=" "><subfield code="a">nicht spezifiziert</subfield><subfield code="b">zu</subfield><subfield code="2">rdacarrier</subfield></datafield><datafield tag="520" ind1=" " ind2=" "><subfield code="a">In this paper, we first extended the semicontinuous hidden Markov model to incorporate multiple code-books. The robustness of the semicontinuous output probability is enhanced by the combination of multiple codewords and multiple codebooks. In addition, we compared the semicontinuous model with the continuous mixture model and the discrete model in a large-vocabulary speaker-independent continuous speech recognition (DARPA resource management) task. The model assumption and parameter size issues are addressed in particular through these experiments. When the acoustic parameters are not well modelled by the continuous probability density, the model assumption problems may cause the recognition accuracy of the semicontinuous model or the continuous mixture model to be inferior to the discrete model. We also found that the SCHMM can have a large number of free parameters in comparison with the discrete HMM because of its smoothing ability. With explicit male and female clustered models and for conditional feature sets, we were able to reduce the error rate of discrete-model-based SPHINX by more than 20%.</subfield></datafield><datafield tag="533" ind1=" " ind2=" "><subfield code="f">Elsevier Journal Backfiles on ScienceDirect 1907 - 2002</subfield></datafield><datafield tag="700" ind1="1" ind2=" "><subfield code="a">Huang, X.D.</subfield><subfield code="4">oth</subfield></datafield><datafield tag="700" ind1="1" ind2=" "><subfield code="a">Hon, H.W.</subfield><subfield code="4">oth</subfield></datafield><datafield tag="700" ind1="1" ind2=" "><subfield code="a">Hwang, M.Y.</subfield><subfield code="4">oth</subfield></datafield><datafield tag="700" ind1="1" ind2=" "><subfield code="a">Lee, K.F.</subfield><subfield code="4">oth</subfield></datafield><datafield tag="773" ind1="0" ind2="8"><subfield code="i">in</subfield><subfield code="t">Computer Speech & Language</subfield><subfield code="d">Amsterdam : Elsevier</subfield><subfield code="g">7(1993), 4, Seite 359-368</subfield><subfield code="w">(DE-627)NLEJ177265698</subfield><subfield code="w">(DE-600)1462883-1</subfield><subfield code="x">0885-2308</subfield><subfield code="7">nnns</subfield></datafield><datafield tag="773" ind1="1" ind2="8"><subfield code="g">volume:7</subfield><subfield code="g">year:1993</subfield><subfield code="g">number:4</subfield><subfield code="g">pages:359-368</subfield></datafield><datafield tag="856" ind1="4" ind2="0"><subfield code="u">http://dx.doi.org/10.1006/csla.1993.1019</subfield></datafield><datafield tag="912" ind1=" " ind2=" "><subfield code="a">GBV_USEFLAG_H</subfield></datafield><datafield tag="912" ind1=" " ind2=" "><subfield code="a">ZDB-1-SDJ</subfield></datafield><datafield tag="912" ind1=" " ind2=" "><subfield code="a">GBV_NL_ARTICLE</subfield></datafield><datafield tag="951" ind1=" " ind2=" "><subfield code="a">AR</subfield></datafield><datafield tag="952" ind1=" " ind2=" "><subfield code="d">7</subfield><subfield code="j">1993</subfield><subfield code="e">4</subfield><subfield code="h">359-368</subfield></datafield></record></collection>
|
| score |
7.400669 |